【摘要】 磁记忆信号检测是分析铁磁材料缺陷和缺陷定位的一种有效方法。由于检测环境噪声、缺陷磁记忆信号弱以及缺陷位置重合引起的信号抵消等原因，易导致传感器采集到的是多维混合磁记忆加噪声复杂信号，体现出较强的盲特性。如何从观测信号中提取有效缺陷信号是分析缺陷类型及定位的关键，传统的降噪方法无法从混合磁记忆观测信号中有效分离缺陷特征信号。文章提出基于核函数的后置非线性盲源分离算法对地下埋管磁记忆信号进行分离，通过与经典盲分离算法Fast ICA进行算法性能比较，表明新算法能够较好地从混合信号中提取有效缺陷特征。更多还原

【Abstract】 Magnetic memory signal detection is an effective method for defect analysis and defect location of ferromagnetic materials. Due to the detection environment noise, weak defect magnetic memory signal and signal cancellation caused by defect location overlap, it is easy to cause the sensor to collect multidimensional mixed magnetic memory and noise complex signal, which embodies strong blind characteristics. How to extract effective defect signals from observation signals is the key to analyze defect types and locate defects. Traditional noise reduction methods cannot effectively separate defect characteristic signals from mixed magnetic memory observation signals. In this paper, the separation of magnetic memory signal of underground pipes based on the kernel function post-nonlinear blind source separation algorithm is proposed. The performance comparison with the classical blind source separation algorithm Fast ICA shows that the new algorithm can better extract effective defect features from mixed signals.更多还原